Leah Morris, Philine Lou Bommer and Alpkaan Celik | Minimum Viable Discoveries: what it takes to get research out of the lab
Foresight Institute’s Computation Group
Minimum Viable Discoveries: what it takes to get research out of the lab
Abstract: Between a scientific discovery and its impact lies a translational gap that existing funding mechanisms are poorly structured to address. Grant funding rewards publication and private capital often rewards revenue. The intermediate stage, in which a result becomes something other researchers can use, is largely unfunded. Encode: AI for Science, a fellowship established by Pillar VC in partnership with the Advanced Research + Invention Agency (ARIA) and the Sovereign AI unit in the UK government, was designed to fund this stage directly. The programme is evaluated against a single unit of progress, the Minimum Viable Discovery (MVD): a functional research output, such as a tool, dataset, model or method, adopted by scientists beyond the originating lab. Where a minimum viable product tests market demand, an MVD tests scientific validity and field adoption.
This talk draws on the experience of eighteen AI researchers embedded for twelve months in UK laboratories to characterise what crossing the translational threshold involves in practice. Examples range from an open-source protein-design model now in use across eight research groups to a neural interface validated in human subjects within the fellowship year. A second finding concerns the changing locus of constraint. AI substantially reduced the cost of generating candidate results, while the cost of validating them remained largely unchanged. When fellows were asked to identify their binding constraint, nine cited data access, seven cited verification capacity, one cited compute, and none cited model capability. The evidence suggests that today, translation remains a verification problem, with implications for the infrastructure that should be developed alongside computational resources.
Speakers Bio:
Prior to her role at Pillar VC, Leah served as a Senior Director at Radical Ventures, where she oversaw the firm’s responsible AI strategy and managed research partnerships. Her background includes economic AI research at the University of Toronto and experience in global health and security with the United Nations and through nonprofit work in Jamaica and South Africa. Leah holds an MBA from the University of Toronto, a Master of Global Affairs from the Munk School of Global Affairs and Public Policy, and dual Bachelor’s degrees in International Development Studies (IDS) and Environment, Sustainability, and Society (ESS) from the University of King’s College.
Philine is completing a PhD in Explainable AI (XAI) for Climate Science from TU Berlin. She built QuantusXClimate, the first-ever tutorial on XAI evaluation, comparison, and selection for climate teams. She led a projects developing frontier XAI methods and building S2S weather forecasting models, submitting to top-tier ML conferences and interdisciplinary journals.
Alpkaan holds a BA in Applied Math from Harvard, where he was the Head Teaching Fellow. He worked in technology and AI for 10 years, most recently at Meta where he was a Machine Learning Engineer in the Health Tech team. Previously, he led Data Science and Machine Learning teams in various startups. Alpkaan has also spent time as an EIR at Entrepreneur First.
A group of scientists, engineers, and entrepreneurs in computer science, ML, cryptography, and related fields who leverage those technologies to improve voluntary cooperation across humans, and ultimately AIs.
Subscribe to our newsletter
Nominate a seminar presenter/topic
Share this application with colleagues who’d like to join
Our book Gaming the Future: Technologies for Intelligent Voluntary Cooperation is now online on Substack.
Zoom link: https://us02web.zoom.us/j/81623367983